ABSTRACT
This chapter considers a design that allows for the comparison of ≥ treatments that incorporate two blocking factors into the experiment. It is easy to reparametrize the less than full rank traditional Latin square design (LSD) model into a full rank regression model. One can then use regression techniques to analyze data from an LSD. The LSD is often used in drug studies. In some problems a subject is given several drugs, and we might be concerned about possible order effects of the drugs. An LSD can be used to help "even out" the order effect of the drug. An extension of a Latin square to four factors can be accomplished by using a Graeco–Latin square design. The Latin square is very useful in developing effective crossover designs. It is easy to reparametrize the less than full rank traditional LSD model into a full rank regression model.
